The present invention relates to a technique for generating an engineering shape model using a 3-D digitizer, and more particularly, to a shape model generation technique capable of generating a high-precision shape model using a volume scanner such as an X-ray CT as the 3-D digitizer.
In digital engineering using CAD, a technique using the shape of an actual object is collectively called “reverse engineering” in general. The use of the description “reverse” is attributable to the circumstance under which there has been no systematic method for obtaining CAD data from the shape of an actual object so far and the actual object has to be necessarily generated from CAD data. However, prime importance is being placed on reverse engineering for the purpose of making most of the shapes of natural objects such as living body and feeding back the expertise of skilled technicians.
The core of reverse engineering is a “3-D digitizer” which measures a 3-D shape of an object and converts the 3-D shape to 3-D data. As the 3-D digitizer, a probe-type 3-D measuring instrument (CMM: Coordinate Measuring Machine) and optical digitizer have often been used so far. However, X-ray CT capable of converting an inner shape as well as density distribution of an object to data is recently attracting attention. The reverse engineering making full use of an X-ray CT is described more specifically in “Concurrent Engineering using an X-Ray CT and RP” (Katsutoshi Sato, Taro Takagi, Shigeru Idemi, 15th Rapid Prototyping Symposium material (1998)), for example.
A probe-type 3-D measuring instrument and optical digitizer are devices which output “point cloud data.” The “point cloud data” is a data format which describes a set of many points (point cloud) which exist on the surface of an actual object. These devices convert only the surface of the object to data, and therefore they are generally called “surface scanners.” The point cloud data as is cannot be used in CAD. However, if neighboring points are linked with one another and polygon data is generated, it is possible to divide the 3-D space into the inside and outside of the object and use the polygon data as a “shape model” which contributes to engineering. The work of generating the polygon data from point cloud data is called “mesh generation.”
On the other hand, an X-ray CT is a device which outputs “bitmap data.” The “bitmap data” is a data format for describing a shape using spatially arranged unit elements called “cells.” Since the X-ray CT can convert the inside of an object to data, it is generally called a “volume scanner.” The individual cells have a value called a “cell value” which represents an X-ray absorption coefficient of the component material at that point of the object. This value is substantially proportional to the density of the material. Therefore, if a cell having a cell value greater than a predetermined threshold is extracted, it is possible to know a 3-D shape of the solid part of the object. Here, the “cell” is generally used in the case of two-dimensional bitmap data, while in the case of three-dimensional bitmap data, the corresponding element is normally called a “boxel,” but in the present specification, “boxel” and “cell” all together may be called “cell.”
The bitmap data includes a great amount of information, but there is little compatibility between bitmap data and CAD data. Therefore, the bitmap data as is cannot be used as a shape model. However, if a cell whose cell value matches a threshold is extracted from the bitmap data, it is possible to obtain point cloud data about the surface of the object based on the coordinates. Once the point cloud data is obtained, it is possible to generate a shape model using mesh generation. This method is the one conventionally used to generate a shape model from X-ray CT data.
The threshold of a cell value is generally an average of a cell value corresponding to an internal material (material in the solid part of an object, for example, aluminum) of the object and a cell value corresponding to an external medium (medium which surrounds the solid part of the object from the outside, for example, air). It is a general practice to carry out an interpolation process between cells instead of extracting cells whose cell value matches a threshold, but the underlying idea is the same.
Examples of techniques related to the background of the present invention are not limited to those described above but also include “3-D Dimension Measuring Apparatus and 3-D Dimension Measuring Method” disclosed in Japanese Patent No. 3431022.
The above described conventional techniques using X-ray CT data have several problems. One of them is that when the thickness of the solid part of an object is smaller than a predetermined value, it is not possible to obtain point cloud data or generate a shape model. X-ray CT data generally includes certain defocusing. An extension of defocusing (defocusing expansion) is normally 2 to 5 cells (the cell size is generally on the order of 100 to 500 μ). Because of this defocusing, if the thickness is equivalent to the defocusing expansion or smaller than the defocusing expansion, the cell value even in a medial part (medial line part) of the thin part does not reach the original value of the material of the part. When the thickness is extremely small, the cell value even does not reach the threshold, and therefore the fact per se that the object exists in that part may be overlooked.
Another problem is that when an object made of a plurality of different materials is handled, it is not possible to model the plurality of different materials simultaneously and accurately. This is attributable to the fact that there are a plurality of internal materials and it is therefore not possible to determine a threshold of an appropriate cell value.
A further problem is that it is not possible to make full use of shape features (features) of the object. With the conventional technique which converts bitmap data to point cloud data, many features are lost making it no longer possible to use a technique which statistically eliminates measurement errors making full use of the features.
Against the above described background, the present invention relates to generation of a shape model from shape description data such as X-ray CT data and it is a first object of the present invention to provide a more user-friendly shape model generation method capable of eliminating the above described various problems, it is a second object to provide a computer program implementing the method and it is a third object to provide a shape model system for executing the above described method.